Background
The BRCA1-associated RING domain protein-1 (BARD1) was initially reported as a BRCA1-interacting protein by Wu et al. in 1996 [
1]. The BRCA1 and BARD1 proteins show high structural homology, as they share
N-terminal RING finger domains and BRCA1 C-terminal (BRCT) domains. Both proteins can form homodimers via their
N-terminal RING finger domains [
2,
3] but preferentially form more stable heterodimers involving amino acid residues 1–109 of the BRCA1 protein and amino acid residues 26–119 of the BARD1 protein [
4]. The interaction between BARD1 and BRCA1 promotes tumor suppressor functions by acting in double-strand break repair and apoptosis initiation.
While the role of the
BRCA1 gene (MIM *113705) in breast cancer (BC) and ovarian cancer (OC) predisposition is well established [
5], the role of
BARD1 (MIM *601593) in BC/OC predisposition remains elusive. Several case-control studies have investigated the association between deleterious germline variants in
BARD1 and the risk of developing female BC. Slavin et al. identified deleterious
BARD1 variants in 7 of 2134
BRCA1/2-negative familial BC patients (carrier frequency = 0.33%) and reported
BARD1 as a moderate-risk BC predisposition gene with an odds ratio (OR) of 3.18 (95% confidence interval [CI] = 1.34–7.36;
P = 0.012) [
6]. The considerably larger investigation of 28,536 BC patients of European ancestry by Couch et al. [
7] revealed a carrier frequency of 0.18% (52/28,536) in BC patients and an OR of 2.16 (95% CI = 1.31–3.63;
P = 0.00226). In contrast, however, recent studies by Lu et al. and Castéra et al. encompassing 9639 and 3667 patients with BC, respectively, did not show a significant association of deleterious
BARD1 variants with overall BC risk [
8,
9]. Studies investigating the association of deleterious
BARD1 germline variants with OC risk, likewise, showed contradictory results. Norquist et al. identified protein-truncating germline variants in 4 of 1915 OC patients unselected for age or family history and in 18 of 36,276 control individuals, resulting in an OR of 4.2 (95% CI = 1.4–12.5;
P = 0.02) [
10]. In contrast, Ramus et al. were unable to demonstrate a significant association with OC in their study of 3261 unselected patients with epithelial OC and 3449 control individuals (4/3261, carrier frequency = 0.12%; 2/3449, carrier frequency = 0.06%;
P = 0.39) [
11]. Lilyquist et al. found deleterious
BARD1 germline variants in 8 of 6294 OC patients (carrier frequency = 0.13%) and calculated a nonsignificant risk ratio of 1.28 (95% CI = 0.55–2.51;
P = 0.59) for OC [
12]. Taken together, the role of deleterious
BARD1 germline variants in BC/OC predisposition remains unclear. In this study, we investigated the prevalence of deleterious
BARD1 germline variants in a sample of 4469 familial BC and 451 familial OC index patients of the German Consortium for Hereditary Breast and Ovarian Cancer (GC-HBOC) and 2767 geographically matched female controls (GMCs).
Results
In our study sample of 4469 familial BC index patients, 23 patients carried heterozygous germline LoF variants in
BARD1, resulting in a carrier frequency of 0.51% (Table
1). One
BARD1-mutated BC index patient additionally carried a heterozygous germline LoF variant in the
CHEK2 gene (patient 5; c.902del, p.Glu301Glyfs*; Additional file
1: Table S2). The remaining 22
BARD1-mutated index patients tested negative for pathogenic variants in further BC/OC predisposition genes (
ATM,
BRCA1,
BRCA2,
CDH1,
CHEK2,
PALB2,
RAD51C,
RAD51D, and
TP53). Information regarding the hormone receptor (estrogen receptor [ER]/progesterone receptor [PR]) and human epidermal growth factor receptor 2 (HER2) status of the tumor was available for 20/23
BARD1-mutated index patients with BC (Additional file
1: Table S2). Most
BARD1-mutated index patients with BC developed hormone receptor-positive (ER-positive: 15/20; PR-positive: 11/20) and HER2-negative tumors (20/20). A triple-negative tumor phenotype was reported for 4 of 20
BARD1-mutated index patients with BC (Additional file
1: Table S2).
The carrier frequency observed in 4469 familial index patients with BC was elevated compared with the carrier frequencies observed in control datasets, which ranged from 0.07% (GMCs) to 0.11% (FLOSSIES) (Table
1). The comparison of carrier frequencies in the study sample of 4469 familial index patients with BC (23/4469, carrier frequency = 0.51%) and all control individuals (36/37,265, carrier frequency = 0.10%) revealed an OR of 5.35 (95% CI = 3.17–9.04;
P < 0.00001) (Table
1). The subgroup of index patients with BC and heterozygous germline LoF variants in
BARD1 showed a younger mean AAD of BC (42.3 years; range 24–60 years) compared with the overall sample of index patients with BC (48.6 years; range 17–92 years), with differences reaching levels of significance (
P = 0.00347; Student’s
t test). When comparing LoF variant prevalence in the subgroup of index patients with BC and an AAD < 40 years and all control individuals, an OR of 12.04 (95% CI = 5.78–25.08;
P < 0.00001) was observed (Table
1). An OR of 7.43 (95% CI = 4.26–12.98;
P < 0.00001) was observed when stratified for an AAD < 50 years. Heterozygous germline LoF variants in
BARD1 were not significantly associated with BC in the subgroup of 1807 BC index patients with an AAD ≥ 50 years, although the ORs were marginally elevated (Table
1). All heterozygous germline LoF variants in
BARD1 identified in patients with BC and in control individuals are listed in the supplements (Additional file
1: Table S3).
Data on proven pathogenic
BARD1 missense variants are currently lacking [
17‐
22]. To examine the potential association of missense variants in
BARD1 with BC risk, we focused on potentially damaging rare missense variants (MAF < 0.1%), which were predicted to be damaging by the SIFT and MutationTaster algorithms. The carrier frequency of potentially damaging rare
BARD1 missense variants was 0.18% for all control individuals (66/37,265; Additional file
1: Table S4). Rare
BARD1 missense variants predicted to be damaging by both tools were significantly more prevalent in index patients with BC compared with control individuals (17/4469, carrier frequency = 0.38%; OR = 2.15; 95% CI = 1.26–3.67;
P = 0.00723; Additional file
1: Table S4). A slightly elevated association was observed for potentially damaging rare
BARD1 missense variants affecting the two BRCT domains spanning the amino acid residues 560–653 and 667–777 (9/4469, carrier frequency = 0.20%; OR = 2.42; 95% CI = 1.15–5.09;
P = 0.03398; Additional file
1: Table S4).
In summary,
BARD1 appears to be a risk gene for early-onset familial BC. To avoid a recruitment bias by OC, we next stratified the study sample according to family history. In the subgroup of 3651 index patients with BC and without an OC family history (mean AAD 48.3 years; range 19–91 years), 19 patients carried heterozygous germline LoF variants in
BARD1, resulting in a carrier frequency of 0.52% and an OR of 5.41 (95% CI = 3.10–9.44;
P < 0.00001) compared with all control individuals (Table
1). In the subgroup of 818 index patients with BC and at least one relative with OC (mean AAD 50.1 years; range 17–92 years), 4 index patients carried heterozygous germline LoF variants in
BARD1 (carrier frequency = 0.49%) and an OR of 5.08 (95% CI = 1.81–14.31;
P = 0.01046) compared with all control individuals. Thus, an OC family history did not affect the prevalence of
BARD1 LoF variants. The analysis of 451 familial index patients with OC (mean AAD 53.4 years; range 18–85 years) did not reveal heterozygous germline LoF variants in
BARD1 (Table
1), and none of the patients with OC carried potentially damaging rare
BARD1 missense variants.
Discussion
We did not observe evidence that deleterious
BARD1 gene variants predispose for OC. LoF germline variants in
BARD1 could neither be detected in 451 familial OC index patients investigated in this study nor in our previously published analysis of 523 consecutive OC patients enrolled in the observational AGO-TR1 study [
23]. Our data are in line with the data provided by Ramus et al. [
11] and the largest investigation to date of 6294 OC cases by Lilyquist et al. [
12], which showed a similar
BARD1 mutation prevalence in OC patients and controls. The weak association previously described by Norquist et al. (
P = 0.02) [
10] was based on the identification of 4
BARD1-mutated individuals in a study sample of 1915 unselected OC patients. Of note, Norquist et al. indicated that these results should be interpreted with some caution as 2 of the 4
BARD1 mutation carriers also had mutations in
BRCA1 [
10]. Overall, it appears likely that deleterious germline
BARD1 variants do not predispose for OC.
In study samples selected for (positive) cancer family history, the prevalence of deleterious variants in established risk genes is generally higher than in unselected cases. In our study focusing on 4469 index patients with familial BC, we demonstrate a significant association of heterozygous germline LoF variants in
BARD1 and overall BC (OR = 5.35; 95% CI = 3.17–9.04;
P < 0.00001). This association is comparable with that described by Slavin et al. (OR = 3.18; 95% CI = 1.34–7.36;
P = 0.012), a study that also focused on index cases with familial BC. In study samples unselected for family history, the observed ORs were lower (e.g., Couch et al.: OR = 2.16; 95% CI = 1.31–3.63;
P = 0.00226) [
7] and even nonsignificant (e.g., Castéra et al.: OR = 2.00; 95% CI = 0.74–4.10) [
8,
9]. Thus, it appears worthwhile to stratify study results by family history and possibly AAD, as shown in the current study. We demonstrate a significant association of heterozygous germline LoF variants in
BARD1 and the risk of early-onset BC (Table
1), a finding which may have important implications for the clinical management of women carrying pathogenic
BARD1 variants. Due to the pronounced association with early-onset BC (AAD < 40 years: OR = 12.04; AAD < 50 years: OR = 7.43), we suggest that
BARD1 should be included in multigene panels for BC risk assessment and, due to the comparatively young AAD of BC observed, intensified BC surveillance programs should be offered to women carrying pathogenic variants in
BARD1.